diff --git a/litellm/litellm_core_utils/get_supported_openai_params.py b/litellm/litellm_core_utils/get_supported_openai_params.py
--- a/litellm/litellm_core_utils/get_supported_openai_params.py
+++ b/litellm/litellm_core_utils/get_supported_openai_params.py
@@ -11,6 +11,7 @@
     request_type: Literal[
         "chat_completion", "embeddings", "transcription"
     ] = "chat_completion",
+    base_model: Optional[str] = None,
 ) -> Optional[list]:
     """
     Returns the supported openai params for a given model + provider
@@ -20,6 +21,11 @@
     get_supported_openai_params(model="anthropic.claude-3", custom_llm_provider="bedrock")
     ```
 
+    Args:
+        base_model: For Azure, the true underlying model (e.g. ``"azure/gpt-5.2"``)
+            when the deployment name differs. Used for model-type detection so that
+            non-standard deployment names route to the correct config.
+
     Returns:
     - List if custom_llm_provider is mapped
     - None if unmapped
@@ -32,17 +38,21 @@
 
     if custom_llm_provider in LlmProvidersSet:
         provider_config = litellm.ProviderConfigManager.get_provider_chat_config(
-            model=model, provider=LlmProviders(custom_llm_provider)
+            model=model,
+            provider=LlmProviders(custom_llm_provider),
+            base_model=base_model,
         )
     elif custom_llm_provider.split("/")[0] in LlmProvidersSet:
         provider_config = litellm.ProviderConfigManager.get_provider_chat_config(
-            model=model, provider=LlmProviders(custom_llm_provider.split("/")[0])
+            model=model,
+            provider=LlmProviders(custom_llm_provider.split("/")[0]),
+            base_model=base_model,
         )
     else:
         provider_config = None
 
     if provider_config and request_type == "chat_completion":
-        return provider_config.get_supported_openai_params(model=model)
+        return provider_config.get_supported_openai_params(model=base_model or model)
 
     if custom_llm_provider == "bedrock":
         return litellm.AmazonConverseConfig().get_supported_openai_params(model=model)
@@ -130,16 +140,23 @@
                 model=model
             )
     elif custom_llm_provider == "azure":
-        if litellm.AzureOpenAIO1Config().is_o_series_model(model=model):
+        _azure_detection_model = base_model or model
+        if litellm.AzureOpenAIO1Config().is_o_series_model(
+            model=_azure_detection_model
+        ):
             return litellm.AzureOpenAIO1Config().get_supported_openai_params(
-                model=model
+                model=_azure_detection_model
             )
-        elif litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(model=model):
+        elif litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(
+            model=_azure_detection_model
+        ):
             return litellm.AzureOpenAIGPT5Config().get_supported_openai_params(
-                model=model
+                model=_azure_detection_model
             )
         else:
-            return litellm.AzureOpenAIConfig().get_supported_openai_params(model=model)
+            return litellm.AzureOpenAIConfig().get_supported_openai_params(
+                model=_azure_detection_model
+            )
     elif custom_llm_provider == "openrouter":
         return litellm.OpenrouterConfig().get_supported_openai_params(model=model)
     elif custom_llm_provider == "vercel_ai_gateway":

diff --git a/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py b/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py
--- a/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py
+++ b/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py
@@ -1476,7 +1476,7 @@
         for choice in choices:
             if choice.delta.content is not None and len(choice.delta.content) > 0:
                 text += choice.delta.content
-            if choice.delta.tool_calls is not None:
+            if choice.delta.tool_calls:
                 partial_json = ""
                 for tool in choice.delta.tool_calls:
                     if (

diff --git a/litellm/llms/azure/azure.py b/litellm/llms/azure/azure.py
--- a/litellm/llms/azure/azure.py
+++ b/litellm/llms/azure/azure.py
@@ -239,7 +239,9 @@
                 )
 
                 data = {"model": None, "messages": messages, **optional_params}
-            elif litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(model=model):
+            elif litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(
+                model=litellm_params.get("base_model") or model
+            ):
                 data = litellm.AzureOpenAIGPT5Config().transform_request(
                     model=model,
                     messages=messages,

diff --git a/litellm/main.py b/litellm/main.py
--- a/litellm/main.py
+++ b/litellm/main.py
@@ -1491,7 +1491,9 @@
             provider.value for provider in LlmProviders
         ]:
             provider_config = ProviderConfigManager.get_provider_chat_config(
-                model=model, provider=LlmProviders(custom_llm_provider)
+                model=model,
+                provider=LlmProviders(custom_llm_provider),
+                base_model=base_model,
             )
 
         if provider_config is not None:
@@ -1550,6 +1552,7 @@
             "safety_identifier": safety_identifier,
             "service_tier": service_tier,
             "allowed_openai_params": kwargs.get("allowed_openai_params"),
+            "base_model": base_model,
         }
         optional_params = get_optional_params(
             **optional_param_args, **non_default_params
@@ -1713,7 +1716,7 @@
             and OpenAIGPT5Config.is_model_gpt_5_model(model)
         ) or (
             custom_llm_provider == "azure"
-            and litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(model)
+            and litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(base_model or model)
         ):
             optional_params, _ = strip_reasoning_summary_aliases_from_optional_params(
                 optional_params
@@ -1766,7 +1769,12 @@
             if max_retries is not None:
                 optional_params["max_retries"] = max_retries
 
-            if litellm.AzureOpenAIO1Config().is_o_series_model(model=model):
+            # Use base_model (the true underlying model) for model-type
+            # detection when the deployment name differs from the model name.
+            _azure_detection_model = base_model or model
+            if litellm.AzureOpenAIO1Config().is_o_series_model(
+                model=_azure_detection_model
+            ):
                 ## LOAD CONFIG - if set
                 config = litellm.AzureOpenAIO1Config.get_config()
                 for k, v in config.items():

diff --git a/litellm/utils.py b/litellm/utils.py
--- a/litellm/utils.py
+++ b/litellm/utils.py
@@ -4019,16 +4019,23 @@
     thinking: Optional[AnthropicThinkingParam] = None,
     web_search_options: Optional[OpenAIWebSearchOptions] = None,
     safety_identifier: Optional[str] = None,
+    base_model: Optional[str] = None,
     **kwargs,
 ):
     passed_params = locals().copy()
     special_params = passed_params.pop("kwargs")
+    # Remove base_model from passed_params so it doesn't interfere with
+    # non_default_params / _check_valid_arg — it's a routing hint, not an
+    # OpenAI param.
+    passed_params.pop("base_model", None)
     provider_config: Optional[BaseConfig] = None
     if custom_llm_provider is not None and custom_llm_provider in [
         provider.value for provider in LlmProviders
     ]:
         provider_config = ProviderConfigManager.get_provider_chat_config(
-            model=model, provider=LlmProviders(custom_llm_provider)
+            model=model,
+            provider=LlmProviders(custom_llm_provider),
+            base_model=base_model,
         )
     non_default_params = pre_process_non_default_params(
         passed_params=passed_params,
@@ -4091,7 +4098,7 @@
         sys.modules[__name__], "get_supported_openai_params"
     )
     supported_params = get_supported_openai_params(
-        model=model, custom_llm_provider=custom_llm_provider
+        model=model, custom_llm_provider=custom_llm_provider, base_model=base_model
     )
     if supported_params is None:
         supported_params = get_supported_openai_params(
@@ -4702,22 +4709,27 @@
             ),
         )
     elif custom_llm_provider == "azure":
-        if litellm.AzureOpenAIO1Config().is_o_series_model(model=model):
+        _azure_detection_model = base_model or model
+        if litellm.AzureOpenAIO1Config().is_o_series_model(
+            model=_azure_detection_model
+        ):
             optional_params = litellm.AzureOpenAIO1Config().map_openai_params(
                 non_default_params=non_default_params,
                 optional_params=optional_params,
-                model=model,
+                model=_azure_detection_model,
                 drop_params=(
                     drop_params
                     if drop_params is not None and isinstance(drop_params, bool)
                     else False
                 ),
             )
-        elif litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(model=model):
+        elif litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(
+            model=_azure_detection_model
+        ):
             optional_params = litellm.AzureOpenAIGPT5Config().map_openai_params(
                 non_default_params=non_default_params,
                 optional_params=optional_params,
-                model=model,
+                model=_azure_detection_model,
                 drop_params=(
                     drop_params
                     if drop_params is not None and isinstance(drop_params, bool)
@@ -4739,7 +4751,7 @@
             optional_params = litellm.AzureOpenAIConfig().map_openai_params(
                 non_default_params=non_default_params,
                 optional_params=optional_params,
-                model=model,
+                model=_azure_detection_model,
                 api_version=api_version,  # type: ignore
                 drop_params=(
                     drop_params
@@ -8124,10 +8136,8 @@
             # Format: (factory_function, needs_model_parameter: bool)
             LlmProviders.OPENAI: (lambda: litellm.OpenAIGPTConfig(), False),
             LlmProviders.ANTHROPIC: (lambda: litellm.AnthropicConfig(), False),
-            LlmProviders.AZURE: (
-                lambda model: ProviderConfigManager._get_azure_config(model),
-                True,
-            ),
+            # AZURE is handled as a special case in get_provider_chat_config()
+            # so that base_model can be threaded through for model-type detection.
             LlmProviders.AZURE_AI: (
                 lambda model: ProviderConfigManager._get_azure_ai_config(model),
                 True,
@@ -8267,11 +8277,19 @@
         }
 
     @staticmethod
-    def _get_azure_config(model: str) -> BaseConfig:
-        """Get Azure config based on model type."""
-        if litellm.AzureOpenAIO1Config().is_o_series_model(model=model):
+    def _get_azure_config(model: str, base_model: Optional[str] = None) -> BaseConfig:
+        """Get Azure config based on model type.
+
+        When *base_model* is provided (e.g. ``"azure/gpt-5.2"``), it is used
+        for model-type detection instead of *model* (the deployment name).
+        This allows non-standard deployment names like ``"azure/foo"`` to be
+        routed through the correct config when the user specifies the true
+        underlying model via ``base_model``.
+        """
+        detection_model = base_model or model
+        if litellm.AzureOpenAIO1Config().is_o_series_model(model=detection_model):
             return litellm.AzureOpenAIO1Config()
-        if litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(model=model):
+        if litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(model=detection_model):
             return litellm.AzureOpenAIGPT5Config()
         return litellm.AzureOpenAIConfig()
 
@@ -8329,13 +8347,18 @@
 
     @staticmethod
     def get_provider_chat_config(  # noqa: PLR0915
-        model: str, provider: LlmProviders
+        model: str,
+        provider: LlmProviders,
+        base_model: Optional[str] = None,
     ) -> Optional[BaseConfig]:
         """
         Returns the provider config for a given provider.
 
         Uses O(1) dictionary lookup for fast provider resolution.
         Python classes take priority over JSON (they have custom overrides).
+
+        For Azure, *base_model* (when set) drives model-type detection so that
+        non-standard deployment names still route to the correct config.
         """
         # Handle OpenAI special cases (O-series and GPT-5 models)
         if provider == LlmProviders.OPENAI:
@@ -8344,6 +8367,12 @@
             if litellm.OpenAIGPT5Config.is_model_gpt_5_model(model=model):
                 return litellm.OpenAIGPT5Config()
 
+        # Handle Azure before the generic map so base_model can be threaded through
+        if provider == LlmProviders.AZURE:
+            return ProviderConfigManager._get_azure_config(
+                model=model, base_model=base_model
+            )
+
         # Initialize provider config map lazily (avoids circular imports)
         if ProviderConfigManager._PROVIDER_CONFIG_MAP is None:
             ProviderConfigManager._PROVIDER_CONFIG_MAP = (

diff --git a/tests/test_litellm/llms/anthropic/experimental_pass_through/adapters/test_anthropic_experimental_pass_through_adapters_transformation.py b/tests/test_litellm/llms/anthropic/experimental_pass_through/adapters/test_anthropic_experimental_pass_through_adapters_transformation.py
--- a/tests/test_litellm/llms/anthropic/experimental_pass_through/adapters/test_anthropic_experimental_pass_through_adapters_transformation.py
+++ b/tests/test_litellm/llms/anthropic/experimental_pass_through/adapters/test_anthropic_experimental_pass_through_adapters_transformation.py
@@ -1203,6 +1203,51 @@
     assert content_block_start["id"] == "toolu_bdrk_013xRVejhv3ybmLEGCoZib2b"
 
 
+def test_streaming_chunk_with_text_and_empty_tool_calls_returns_text_delta():
+    """
+    Some OpenAI-compatible providers emit `tool_calls: []` on regular text chunks.
+
+    Empty tool_calls should be treated as no tool call so the Anthropic adapter
+    does not shadow text with an empty input_json_delta.
+    """
+    choices = [
+        StreamingChoices(
+            finish_reason=None,
+            index=0,
+            delta=Delta(
+                provider_specific_fields=None,
+                content="Hello from vLLM",
+                role="assistant",
+                function_call=None,
+                tool_calls=[],
+                audio=None,
+            ),
+            logprobs=None,
+        )
+    ]
+
+    adapter = LiteLLMAnthropicMessagesAdapter()
+
+    (
+        type_of_content,
+        content_block_delta,
+    ) = adapter._translate_streaming_openai_chunk_to_anthropic(choices=choices)
+
+    assert type_of_content == "text_delta"
+    assert content_block_delta["type"] == "text_delta"
+    assert content_block_delta["text"] == "Hello from vLLM"
+
+    (
+        block_type,
+        content_block_start,
+    ) = adapter._translate_streaming_openai_chunk_to_anthropic_content_block(
+        choices=choices
+    )
+
+    assert block_type == "text"
+    assert content_block_start == {"type": "text", "text": ""}
+
+
 # ============================================================================
 # Cache Control Transformation Tests
 # ============================================================================

diff --git a/tests/test_litellm/llms/azure/chat/test_azure_base_model_routing.py b/tests/test_litellm/llms/azure/chat/test_azure_base_model_routing.py
new file mode 100644
--- /dev/null
+++ b/tests/test_litellm/llms/azure/chat/test_azure_base_model_routing.py
@@ -1,0 +1,274 @@
+"""Tests for decoupling Azure deployment IDs from underlying model names.
+
+When users name their Azure deployment something non-standard (e.g. "my-deployment-id"),
+setting ``base_model`` should drive model-type detection (o-series, gpt-5,
+etc.) so the correct config, supported params, and param mapping are used.
+"""
+
+import pytest
+
+import litellm
+from litellm.llms.azure.chat.gpt_5_transformation import AzureOpenAIGPT5Config
+from litellm.llms.azure.chat.o_series_transformation import AzureOpenAIO1Config
+from litellm.utils import ProviderConfigManager, get_optional_params
+
+
+# ---------------------------------------------------------------------------
+# _get_azure_config — routes to the correct config based on base_model
+# ---------------------------------------------------------------------------
+class TestGetAzureConfigWithBaseModel:
+    """ProviderConfigManager._get_azure_config should use base_model for detection."""
+
+    def test_should_return_gpt5_config_when_base_model_is_gpt5(self):
+        config = ProviderConfigManager._get_azure_config(
+            model="my-deployment-id", base_model="azure/gpt-5.2"
+        )
+        assert isinstance(config, AzureOpenAIGPT5Config)
+
+    def test_should_return_o_series_config_when_base_model_is_o_series(self):
+        config = ProviderConfigManager._get_azure_config(
+            model="my-deployment-id", base_model="azure/o4-mini"
+        )
+        assert isinstance(config, AzureOpenAIO1Config)
+
+    def test_should_return_default_config_when_base_model_is_regular(self):
+        config = ProviderConfigManager._get_azure_config(
+            model="my-deployment-id", base_model="azure/gpt-4o"
+        )
+        assert type(config).__name__ == "AzureOpenAIConfig"
+
+    def test_should_fallback_to_model_when_base_model_is_none(self):
+        config = ProviderConfigManager._get_azure_config(
+            model="gpt-5.2", base_model=None
+        )
+        assert isinstance(config, AzureOpenAIGPT5Config)
+
+    def test_should_return_default_config_when_both_are_non_standard(self):
+        config = ProviderConfigManager._get_azure_config(
+            model="my-deployment-id", base_model=None
+        )
+        assert type(config).__name__ == "AzureOpenAIConfig"
+
+
+# ---------------------------------------------------------------------------
+# get_provider_chat_config — threads base_model through for Azure
+# ---------------------------------------------------------------------------
+class TestGetProviderChatConfigWithBaseModel:
+    """get_provider_chat_config should pass base_model to Azure config selection."""
+
+    def test_should_return_gpt5_config_for_custom_deployment_with_base_model(self):
+        from litellm.types.utils import LlmProviders
+
+        config = ProviderConfigManager.get_provider_chat_config(
+            model="my-deployment-id",
+            provider=LlmProviders.AZURE,
+            base_model="azure/gpt-5",
+        )
+        assert isinstance(config, AzureOpenAIGPT5Config)
+
+    def test_should_return_o_series_config_for_custom_deployment_with_base_model(self):
+        from litellm.types.utils import LlmProviders
+
+        config = ProviderConfigManager.get_provider_chat_config(
+            model="my-other-deployment",
+            provider=LlmProviders.AZURE,
+            base_model="azure/o3-mini",
+        )
+        assert isinstance(config, AzureOpenAIO1Config)
+
+
+# ---------------------------------------------------------------------------
+# get_supported_openai_params — base_model drives Azure param detection
+# ---------------------------------------------------------------------------
+class TestGetSupportedOpenAIParamsWithBaseModel:
+    """get_supported_openai_params should use base_model for Azure detection."""
+
+    def test_should_return_gpt5_params_for_custom_deployment_with_gpt5_base_model(
+        self,
+    ):
+        params = litellm.get_supported_openai_params(
+            model="my-deployment-id",
+            custom_llm_provider="azure",
+            base_model="azure/gpt-5",
+        )
+        assert params is not None
+        assert "reasoning_effort" in params
+        # gpt-5 maps max_tokens -> max_completion_tokens, verifying we got GPT-5 config
+        assert "max_completion_tokens" in params
+
+    def test_should_return_o_series_params_for_custom_deployment_with_o_series_base_model(
+        self,
+    ):
+        params = litellm.get_supported_openai_params(
+            model="my-other-deployment",
+            custom_llm_provider="azure",
+            base_model="azure/o4-mini",
+        )
+        assert params is not None
+        assert "reasoning_effort" in params
+
+    def test_should_return_regular_params_when_no_base_model(self):
+        """When base_model is not set and model is non-standard, default Azure config."""
+        params = litellm.get_supported_openai_params(
+            model="my-deployment-id",
+            custom_llm_provider="azure",
+        )
+        assert params is not None
+        # Default Azure config supports temperature
+        assert "temperature" in params
+
+
+# ---------------------------------------------------------------------------
+# get_optional_params — base_model drives Azure param mapping
+# ---------------------------------------------------------------------------
+class TestGetOptionalParamsWithBaseModel:
+    """get_optional_params should use base_model for Azure model-type detection."""
+
+    def test_should_map_max_tokens_for_custom_deployment_with_gpt5_base_model(self):
+        """A non-standard deployment name + gpt-5 base_model should map max_tokens -> max_completion_tokens."""
+        params = get_optional_params(
+            model="my-deployment-id",
+            custom_llm_provider="azure",
+            max_tokens=100,
+            base_model="azure/gpt-5",
+        )
+        assert params.get("max_completion_tokens") == 100
+        assert "max_tokens" not in params
+
+    def test_should_keep_max_tokens_for_custom_deployment_without_base_model(self):
+        """A non-standard deployment name without base_model should use default Azure config."""
+        params = get_optional_params(
+            model="my-deployment-id",
+            custom_llm_provider="azure",
+            max_tokens=100,
+            api_version="2024-05-01-preview",
+        )
+        # Default AzureOpenAIConfig keeps max_tokens as-is (or maps based on api_version)
+        assert "max_tokens" in params or "max_completion_tokens" in params
+
+    def test_should_support_reasoning_effort_for_custom_deployment_with_o_series_base_model(
+        self,
+    ):
+        """A non-standard deployment name + o-series base_model should accept reasoning_effort."""
+        params = get_optional_params(
+            model="my-other-deployment",
+            custom_llm_provider="azure",
+            reasoning_effort="low",
+            base_model="azure/o4-mini",
+        )
+        assert params.get("reasoning_effort") == "low"
+
+    def test_should_reject_temperature_for_custom_deployment_with_gpt5_base_model(
+        self,
+    ):
+        """A non-standard deployment + gpt-5 base_model should reject temperature."""
+        with pytest.raises(litellm.UnsupportedParamsError):
+            get_optional_params(
+                model="my-deployment-id",
+                custom_llm_provider="azure",
+                temperature=0.5,
+                base_model="azure/gpt-5",
+            )
+
+
+# ---------------------------------------------------------------------------
+# Backward compatibility — existing patterns still work
+# ---------------------------------------------------------------------------
+class TestBackwardCompatibility:
+    """Existing model-name-based and prefix-based patterns must keep working."""
+
+    def test_should_detect_gpt5_from_model_name(self):
+        config = ProviderConfigManager._get_azure_config(model="gpt-5.2")
+        assert isinstance(config, AzureOpenAIGPT5Config)
+
+    def test_should_detect_gpt5_from_gpt5_series_prefix(self):
+        config = ProviderConfigManager._get_azure_config(
+            model="gpt5_series/my-deployment"
+        )
+        assert isinstance(config, AzureOpenAIGPT5Config)
+
+    def test_should_detect_o_series_from_model_name(self):
+        config = ProviderConfigManager._get_azure_config(model="o4-mini")
+        assert isinstance(config, AzureOpenAIO1Config)
+
+    def test_should_detect_o_series_from_o_series_prefix(self):
+        config = ProviderConfigManager._get_azure_config(model="o_series/my-deployment")
+        assert isinstance(config, AzureOpenAIO1Config)
+
+    def test_should_handle_gpt5_chat_model_correctly(self):
+        """gpt-5-chat models should NOT be routed to GPT-5 config."""
+        config = ProviderConfigManager._get_azure_config(model="gpt-5-chat")
+        assert type(config).__name__ == "AzureOpenAIConfig"
+
+    def test_base_model_overrides_model_detection(self):
+        """base_model should take priority over model for type detection."""
+        # model looks like o-series, but base_model says gpt-5
+        config = ProviderConfigManager._get_azure_config(
+            model="o3-mini", base_model="azure/gpt-5.2"
+        )
+        assert isinstance(config, AzureOpenAIGPT5Config)
+
+
+# ---------------------------------------------------------------------------
+# Deep config method awareness — base_model flows into config internals
+# ---------------------------------------------------------------------------
+class TestBaseModelFlowsIntoConfigInternals:
+    """base_model should be used by config internal methods (e.g. is_model_gpt_5_2_model)."""
+
+    def test_should_support_logprobs_for_prefixed_deployment_with_gpt52_base_model(
+        self,
+    ):
+        """Deployment 'my-gpt-5.2' with base_model='azure/gpt-5.2' should support logprobs."""
+        params = litellm.get_supported_openai_params(
+            model="gpt5_series/my-gpt-5.2",
+            custom_llm_provider="azure",
+            base_model="azure/gpt-5.2",
+        )
+        assert params is not None
+        assert "logprobs" in params
+        assert "top_logprobs" in params
+
+    def test_should_support_logprobs_for_plain_deployment_with_gpt52_base_model(self):
+        """Deployment 'my-deployment-id' with base_model='azure/gpt-5.2' should support logprobs."""
+        params = litellm.get_supported_openai_params(
+            model="my-deployment-id",
+            custom_llm_provider="azure",
+            base_model="azure/gpt-5.2",
+        )
+        assert params is not None
+        assert "logprobs" in params
+        assert "top_logprobs" in params
+
+    def test_should_not_support_logprobs_for_gpt5_base_model(self):
+        """Deployment with base_model='azure/gpt-5' (not 5.2) should NOT support logprobs."""
+        params = litellm.get_supported_openai_params(
+            model="my-deployment-id",
+            custom_llm_provider="azure",
+            base_model="azure/gpt-5",
+        )
+        assert params is not None
+        assert "logprobs" not in params
+        assert "top_logprobs" not in params
+
+    def test_should_pass_logprobs_through_get_optional_params(self):
+        """logprobs should pass validation in get_optional_params when base_model is gpt-5.2."""
+        params = get_optional_params(
+            model="gpt5_series/my-gpt-5.2",
+            custom_llm_provider="azure",
+            logprobs=True,
+            top_logprobs=5,
+            base_model="azure/gpt-5.2",
+        )
+        assert params.get("logprobs") is True
+        assert params.get("top_logprobs") == 5
+
+    def test_should_map_max_tokens_for_prefixed_deployment_with_gpt5_base_model(self):
+        """my-gpt-5.2 with base_model should correctly map max_tokens -> max_completion_tokens."""
+        params = get_optional_params(
+            model="gpt5_series/my-gpt-5.2",
+            custom_llm_provider="azure",
+            max_tokens=200,
+            base_model="azure/gpt-5.2",
+        )
+        assert params.get("max_completion_tokens") == 200
+        assert "max_tokens" not in params

diff --git a/litellm/litellm_core_utils/get_supported_openai_params.py b/litellm/litellm_core_utils/get_supported_openai_params.py
--- a/litellm/litellm_core_utils/get_supported_openai_params.py
+++ b/litellm/litellm_core_utils/get_supported_openai_params.py
@@ -11,6 +11,7 @@
     request_type: Literal[
         "chat_completion", "embeddings", "transcription"
     ] = "chat_completion",
+    base_model: Optional[str] = None,
 ) -> Optional[list]:
     """
     Returns the supported openai params for a given model + provider
@@ -20,6 +21,11 @@
     get_supported_openai_params(model="anthropic.claude-3", custom_llm_provider="bedrock")
     ```
 
+    Args:
+        base_model: For Azure, the true underlying model (e.g. ``"azure/gpt-5.2"``)
+            when the deployment name differs. Used for model-type detection so that
+            non-standard deployment names route to the correct config.
+
     Returns:
     - List if custom_llm_provider is mapped
     - None if unmapped
@@ -32,17 +38,24 @@
 
     if custom_llm_provider in LlmProvidersSet:
         provider_config = litellm.ProviderConfigManager.get_provider_chat_config(
-            model=model, provider=LlmProviders(custom_llm_provider)
+            model=model,
+            provider=LlmProviders(custom_llm_provider),
+            base_model=base_model,
         )
     elif custom_llm_provider.split("/")[0] in LlmProvidersSet:
         provider_config = litellm.ProviderConfigManager.get_provider_chat_config(
-            model=model, provider=LlmProviders(custom_llm_provider.split("/")[0])
+            model=model,
+            provider=LlmProviders(custom_llm_provider.split("/")[0]),
+            base_model=base_model,
         )
     else:
         provider_config = None
 
     if provider_config and request_type == "chat_completion":
-        return provider_config.get_supported_openai_params(model=model)
+        effective_model = (
+            base_model if custom_llm_provider == "azure" and base_model else model
+        )
+        return provider_config.get_supported_openai_params(model=effective_model)
 
     if custom_llm_provider == "bedrock":
         return litellm.AmazonConverseConfig().get_supported_openai_params(model=model)
@@ -130,13 +143,18 @@
                 model=model
             )
     elif custom_llm_provider == "azure":
-        if litellm.AzureOpenAIO1Config().is_o_series_model(model=model):
+        _azure_detection_model = base_model or model
+        if litellm.AzureOpenAIO1Config().is_o_series_model(
+            model=_azure_detection_model
+        ):
             return litellm.AzureOpenAIO1Config().get_supported_openai_params(
-                model=model
+                model=_azure_detection_model
             )
-        elif litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(model=model):
+        elif litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(
+            model=_azure_detection_model
+        ):
             return litellm.AzureOpenAIGPT5Config().get_supported_openai_params(
-                model=model
+                model=_azure_detection_model
             )
         else:
             return litellm.AzureOpenAIConfig().get_supported_openai_params(model=model)

diff --git a/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py b/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py
--- a/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py
+++ b/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py
@@ -1476,7 +1476,7 @@
         for choice in choices:
             if choice.delta.content is not None and len(choice.delta.content) > 0:
                 text += choice.delta.content
-            if choice.delta.tool_calls is not None:
+            if choice.delta.tool_calls:
                 partial_json = ""
                 for tool in choice.delta.tool_calls:
                     if (

diff --git a/litellm/llms/azure/azure.py b/litellm/llms/azure/azure.py
--- a/litellm/llms/azure/azure.py
+++ b/litellm/llms/azure/azure.py
@@ -239,7 +239,9 @@
                 )
 
                 data = {"model": None, "messages": messages, **optional_params}
-            elif litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(model=model):
+            elif litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(
+                model=litellm_params.get("base_model") or model
+            ):
                 data = litellm.AzureOpenAIGPT5Config().transform_request(
                     model=model,
                     messages=messages,

diff --git a/litellm/llms/openai/responses/transformation.py b/litellm/llms/openai/responses/transformation.py
--- a/litellm/llms/openai/responses/transformation.py
+++ b/litellm/llms/openai/responses/transformation.py
@@ -126,9 +126,21 @@
         litellm_params: GenericLiteLLMParams,
         headers: dict,
     ) -> Dict:
-        """No transform applied since inputs are in OpenAI spec already"""
+        """Strip Anthropic-only `cache_control` markers before sending to OpenAI.
 
+        OpenAI's Responses API rejects unknown fields on input content blocks
+        with HTTP 400 ("Unknown parameter: 'input[0].content[0].cache_control'").
+        Chat Completions strips these in
+        `remove_cache_control_flag_from_messages_and_tools`; mirror that here.
+        """
+
         input = self._validate_input_param(input)
+        tools = response_api_optional_request_params.get("tools")
+        input, tools = self.remove_cache_control_flag_from_input_and_tools(
+            model=model, input=input, tools=tools
+        )
+        if tools is not None:
+            response_api_optional_request_params["tools"] = tools
         final_request_params = dict(
             ResponsesAPIRequestParams(
                 model=model, input=input, **response_api_optional_request_params
@@ -137,6 +149,38 @@
 
         return final_request_params
 
+    def remove_cache_control_flag_from_input_and_tools(
+        self,
+        model: str,  # allows overrides to selectively run this
+        input: Union[str, ResponseInputParam],
+        tools: Optional[List[ALL_RESPONSES_API_TOOL_PARAMS]] = None,
+    ) -> Tuple[
+        Union[str, ResponseInputParam],
+        Optional[List[ALL_RESPONSES_API_TOOL_PARAMS]],
+    ]:
+        """Sibling of `remove_cache_control_flag_from_messages_and_tools` on
+        the chat path. Strips Anthropic-only `cache_control` markers from
+        Responses API input content blocks and tools.
+
+        `filter_value_from_dict` mutates each dict in place, so the same
+        objects are returned.
+        """
+        from litellm.litellm_core_utils.prompt_templates.common_utils import (
+            filter_value_from_dict,
+        )
+
+        if isinstance(input, list):
+            for item in input:
+                if isinstance(item, dict):
+                    filter_value_from_dict(cast(dict, item), "cache_control")
+
+        if tools is not None:
+            for tool in tools:
+                if isinstance(tool, dict):
+                    filter_value_from_dict(cast(dict, tool), "cache_control")
+
+        return input, tools
+
     def _validate_input_param(
         self, input: Union[str, ResponseInputParam]
     ) -> Union[str, ResponseInputParam]:

diff --git a/litellm/main.py b/litellm/main.py
--- a/litellm/main.py
+++ b/litellm/main.py
@@ -1491,7 +1491,9 @@
             provider.value for provider in LlmProviders
         ]:
             provider_config = ProviderConfigManager.get_provider_chat_config(
-                model=model, provider=LlmProviders(custom_llm_provider)
+                model=model,
+                provider=LlmProviders(custom_llm_provider),
+                base_model=base_model,
             )
 
         if provider_config is not None:
@@ -1550,6 +1552,7 @@
             "safety_identifier": safety_identifier,
             "service_tier": service_tier,
             "allowed_openai_params": kwargs.get("allowed_openai_params"),
+            "base_model": base_model,
         }
         optional_params = get_optional_params(
             **optional_param_args, **non_default_params
@@ -1766,7 +1769,12 @@
             if max_retries is not None:
                 optional_params["max_retries"] = max_retries
 
-            if litellm.AzureOpenAIO1Config().is_o_series_model(model=model):
+            # Use base_model (the true underlying model) for model-type
+            # detection when the deployment name differs from the model name.
+            _azure_detection_model = base_model or model
+            if litellm.AzureOpenAIO1Config().is_o_series_model(
+                model=_azure_detection_model
+            ):
                 ## LOAD CONFIG - if set
                 config = litellm.AzureOpenAIO1Config.get_config()
                 for k, v in config.items():

diff --git a/litellm/utils.py b/litellm/utils.py
--- a/litellm/utils.py
+++ b/litellm/utils.py
@@ -2933,6 +2933,13 @@
             except Exception:
                 existing_model = {}
                 model_cost_key = key
+        # ``get_model_info`` returns ``litellm_provider: None`` when the
+        # provider is unknown (e.g. custom deployments registered via
+        # ``Router.add_deployment``). Persisting that None into
+        # ``litellm.model_cost`` causes ``_check_provider_match`` to drop
+        # custom pricing on subsequent cost lookups.
+        if existing_model.get("litellm_provider") is None:
+            existing_model.pop("litellm_provider", None)
         ## override / add new keys to the existing model cost dictionary
         updated_dictionary = _update_dictionary(existing_model, value)
         litellm.model_cost.setdefault(model_cost_key, {}).update(updated_dictionary)
@@ -4019,16 +4026,23 @@
     thinking: Optional[AnthropicThinkingParam] = None,
     web_search_options: Optional[OpenAIWebSearchOptions] = None,
     safety_identifier: Optional[str] = None,
+    base_model: Optional[str] = None,
     **kwargs,
 ):
     passed_params = locals().copy()
     special_params = passed_params.pop("kwargs")
+    # Remove base_model from passed_params so it doesn't interfere with
+    # non_default_params / _check_valid_arg — it's a routing hint, not an
+    # OpenAI param.
+    passed_params.pop("base_model", None)
     provider_config: Optional[BaseConfig] = None
     if custom_llm_provider is not None and custom_llm_provider in [
         provider.value for provider in LlmProviders
     ]:
         provider_config = ProviderConfigManager.get_provider_chat_config(
-            model=model, provider=LlmProviders(custom_llm_provider)
+            model=model,
+            provider=LlmProviders(custom_llm_provider),
+            base_model=base_model,
         )
     non_default_params = pre_process_non_default_params(
         passed_params=passed_params,
@@ -4091,7 +4105,7 @@
         sys.modules[__name__], "get_supported_openai_params"
     )
     supported_params = get_supported_openai_params(
-        model=model, custom_llm_provider=custom_llm_provider
+        model=model, custom_llm_provider=custom_llm_provider, base_model=base_model
     )
     if supported_params is None:
         supported_params = get_supported_openai_params(
@@ -4702,22 +4716,27 @@
             ),
         )
     elif custom_llm_provider == "azure":
-        if litellm.AzureOpenAIO1Config().is_o_series_model(model=model):
+        _azure_detection_model = base_model or model
+        if litellm.AzureOpenAIO1Config().is_o_series_model(
+            model=_azure_detection_model
+        ):
             optional_params = litellm.AzureOpenAIO1Config().map_openai_params(
                 non_default_params=non_default_params,
                 optional_params=optional_params,
-                model=model,
+                model=_azure_detection_model,
                 drop_params=(
                     drop_params
                     if drop_params is not None and isinstance(drop_params, bool)
                     else False
                 ),
             )
-        elif litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(model=model):
+        elif litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(
+            model=_azure_detection_model
+        ):
             optional_params = litellm.AzureOpenAIGPT5Config().map_openai_params(
                 non_default_params=non_default_params,
                 optional_params=optional_params,
-                model=model,
+                model=_azure_detection_model,
                 drop_params=(
                     drop_params
                     if drop_params is not None and isinstance(drop_params, bool)
@@ -4739,7 +4758,7 @@
             optional_params = litellm.AzureOpenAIConfig().map_openai_params(
                 non_default_params=non_default_params,
                 optional_params=optional_params,
-                model=model,
+                model=_azure_detection_model,
                 api_version=api_version,  # type: ignore
                 drop_params=(
                     drop_params
@@ -5510,9 +5529,15 @@
 def _check_provider_match(model_info: dict, custom_llm_provider: Optional[str]) -> bool:
     """
     Check if the model info provider matches the custom provider.
+
+    A missing ``litellm_provider`` key and a ``litellm_provider`` set to
+    ``None`` both mean "no specific provider constraint" and are treated
+    as a wildcard match. ``register_model`` may persist ``None`` here via
+    ``get_model_info`` when a deployment is registered without a provider,
+    so normalising the two cases keeps custom pricing applied consistently.
     """
     if custom_llm_provider and (
-        "litellm_provider" in model_info
+        model_info.get("litellm_provider") is not None
         and model_info["litellm_provider"] != custom_llm_provider
     ):
         if custom_llm_provider == "vertex_ai" and model_info[
@@ -8124,10 +8149,8 @@
             # Format: (factory_function, needs_model_parameter: bool)
             LlmProviders.OPENAI: (lambda: litellm.OpenAIGPTConfig(), False),
             LlmProviders.ANTHROPIC: (lambda: litellm.AnthropicConfig(), False),
-            LlmProviders.AZURE: (
-                lambda model: ProviderConfigManager._get_azure_config(model),
-                True,
-            ),
+            # AZURE is handled as a special case in get_provider_chat_config()
+            # so that base_model can be threaded through for model-type detection.
             LlmProviders.AZURE_AI: (
                 lambda model: ProviderConfigManager._get_azure_ai_config(model),
                 True,
@@ -8267,11 +8290,19 @@
         }
 
     @staticmethod
-    def _get_azure_config(model: str) -> BaseConfig:
-        """Get Azure config based on model type."""
-        if litellm.AzureOpenAIO1Config().is_o_series_model(model=model):
+    def _get_azure_config(model: str, base_model: Optional[str] = None) -> BaseConfig:
+        """Get Azure config based on model type.
+
+        When *base_model* is provided (e.g. ``"azure/gpt-5.2"``), it is used
+        for model-type detection instead of *model* (the deployment name).
+        This allows non-standard deployment names like ``"azure/foo"`` to be
+        routed through the correct config when the user specifies the true
+        underlying model via ``base_model``.
+        """
+        detection_model = base_model or model
+        if litellm.AzureOpenAIO1Config().is_o_series_model(model=detection_model):
             return litellm.AzureOpenAIO1Config()
-        if litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(model=model):
+        if litellm.AzureOpenAIGPT5Config.is_model_gpt_5_model(model=detection_model):
             return litellm.AzureOpenAIGPT5Config()
         return litellm.AzureOpenAIConfig()
 
@@ -8329,13 +8360,18 @@
 
     @staticmethod
     def get_provider_chat_config(  # noqa: PLR0915
-        model: str, provider: LlmProviders
+        model: str,
+        provider: LlmProviders,
+        base_model: Optional[str] = None,
     ) -> Optional[BaseConfig]:
         """
         Returns the provider config for a given provider.
 
         Uses O(1) dictionary lookup for fast provider resolution.
         Python classes take priority over JSON (they have custom overrides).
+
+        For Azure, *base_model* (when set) drives model-type detection so that
+        non-standard deployment names still route to the correct config.
         """
         # Handle OpenAI special cases (O-series and GPT-5 models)
         if provider == LlmProviders.OPENAI:
@@ -8344,6 +8380,12 @@
             if litellm.OpenAIGPT5Config.is_model_gpt_5_model(model=model):
                 return litellm.OpenAIGPT5Config()
 
+        # Handle Azure before the generic map so base_model can be threaded through
+        if provider == LlmProviders.AZURE:
+            return ProviderConfigManager._get_azure_config(
+                model=model, base_model=base_model
+            )
+
         # Initialize provider config map lazily (avoids circular imports)
         if ProviderConfigManager._PROVIDER_CONFIG_MAP is None:
             ProviderConfigManager._PROVIDER_CONFIG_MAP = (

diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json
--- a/model_prices_and_context_window.json
+++ b/model_prices_and_context_window.json
@@ -15006,10 +15006,16 @@
         "supports_service_tier": true
     },
     "gemini-3.1-flash-lite": {
-        "cache_read_input_token_cost": 4.5e-08,
-        "cache_read_input_token_cost_per_audio_token": 9e-08,
-        "input_cost_per_audio_token": 9e-07,
-        "input_cost_per_token": 4.5e-07,
+        "cache_read_input_token_cost": 2.5e-08,
+        "cache_read_input_token_cost_batches": 1.25e-08,
+        "cache_read_input_token_cost_flex": 1.25e-08,
+        "cache_read_input_token_cost_per_audio_token": 5e-08,
+        "cache_read_input_token_cost_priority": 4.5e-08,
+        "input_cost_per_audio_token": 5e-07,
+        "input_cost_per_token": 2.5e-07,
+        "input_cost_per_token_batches": 1.25e-07,
+        "input_cost_per_token_flex": 1.25e-07,
+        "input_cost_per_token_priority": 4.5e-07,
         "litellm_provider": "vertex_ai-language-models",
         "max_audio_length_hours": 8.4,
         "max_audio_per_prompt": 1,
@@ -15021,9 +15027,12 @@
         "max_video_length": 1,
         "max_videos_per_prompt": 10,
         "mode": "chat",
-        "output_cost_per_reasoning_token": 2.7e-06,
-        "output_cost_per_token": 2.7e-06,
-        "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models",
+        "output_cost_per_reasoning_token": 1.5e-06,
+        "output_cost_per_token": 1.5e-06,
+        "output_cost_per_token_batches": 7.5e-07,
+        "output_cost_per_token_flex": 7.5e-07,
+        "output_cost_per_token_priority": 2.7e-06,
+        "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-lite",
         "supported_endpoints": [
             "/v1/chat/completions",
             "/v1/completions",
@@ -17128,10 +17137,16 @@
         "supports_service_tier": true
     },
     "gemini/gemini-3.1-flash-lite": {
-        "cache_read_input_token_cost": 4.5e-08,
-        "cache_read_input_token_cost_per_audio_token": 9e-08,
-        "input_cost_per_audio_token": 9e-07,
-        "input_cost_per_token": 4.5e-07,
+        "cache_read_input_token_cost": 2.5e-08,
+        "cache_read_input_token_cost_batches": 1.25e-08,
+        "cache_read_input_token_cost_flex": 1.25e-08,
+        "cache_read_input_token_cost_per_audio_token": 5e-08,
+        "cache_read_input_token_cost_priority": 4.5e-08,
+        "input_cost_per_audio_token": 5e-07,
+        "input_cost_per_token": 2.5e-07,
+        "input_cost_per_token_batches": 1.25e-07,
+        "input_cost_per_token_flex": 1.25e-07,
+        "input_cost_per_token_priority": 4.5e-07,
         "litellm_provider": "gemini",
         "max_audio_length_hours": 8.4,
         "max_audio_per_prompt": 1,
@@ -17143,10 +17158,13 @@
         "max_video_length": 1,
         "max_videos_per_prompt": 10,
         "mode": "chat",
-        "output_cost_per_reasoning_token": 2.7e-06,
-        "output_cost_per_token": 2.7e-06,
+        "output_cost_per_reasoning_token": 1.5e-06,
+        "output_cost_per_token": 1.5e-06,
+        "output_cost_per_token_batches": 7.5e-07,
+        "output_cost_per_token_flex": 7.5e-07,
+        "output_cost_per_token_priority": 2.7e-06,
         "rpm": 15,
-        "source": "https://ai.google.dev/gemini-api/docs/pricing",
+        "source": "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-lite",
         "supported_endpoints": [
             "/v1/chat/completions",
             "/v1/completions",
@@ -33932,10 +33950,16 @@
         "web_search_billing_unit": "per_query"
     },
     "vertex_ai/gemini-3.1-flash-lite": {
-        "cache_read_input_token_cost": 4.5e-08,
-        "cache_read_input_token_cost_per_audio_token": 9e-08,
-        "input_cost_per_audio_token": 9e-07,
-        "input_cost_per_token": 4.5e-07,
+        "cache_read_input_token_cost": 2.5e-08,
+        "cache_read_input_token_cost_batches": 1.25e-08,
+        "cache_read_input_token_cost_flex": 1.25e-08,
+        "cache_read_input_token_cost_per_audio_token": 5e-08,
+        "cache_read_input_token_cost_priority": 4.5e-08,
+        "input_cost_per_audio_token": 5e-07,
+        "input_cost_per_token": 2.5e-07,
+        "input_cost_per_token_batches": 1.25e-07,
+        "input_cost_per_token_flex": 1.25e-07,
+        "input_cost_per_token_priority": 4.5e-07,
         "litellm_provider": "vertex_ai-language-models",
         "max_audio_length_hours": 8.4,
         "max_audio_per_prompt": 1,
@@ -33947,8 +33971,11 @@
         "max_video_length": 1,
         "max_videos_per_prompt": 10,
         "mode": "chat",
-        "output_cost_per_reasoning_token": 2.7e-06,
-        "output_cost_per_token": 2.7e-06,
+        "output_cost_per_reasoning_token": 1.5e-06,
+        "output_cost_per_token": 1.5e-06,
+        "output_cost_per_token_batches": 7.5e-07,
+        "output_cost_per_token_flex": 7.5e-07,
+        "output_cost_per_token_priority": 2.7e-06,
         "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models",
         "supported_endpoints": [
             "/v1/chat/completions",

diff --git a/tests/test_litellm/llms/anthropic/experimental_pass_through/adapters/test_anthropic_experimental_pass_through_adapters_transformation.py b/tests/test_litellm/llms/anthropic/experimental_pass_through/adapters/test_anthropic_experimental_pass_through_adapters_transformation.py
--- a/tests/test_litellm/llms/anthropic/experimental_pass_through/adapters/test_anthropic_experimental_pass_through_adapters_transformation.py
+++ b/tests/test_litellm/llms/anthropic/experimental_pass_through/adapters/test_anthropic_experimental_pass_through_adapters_transformation.py
@@ -1203,6 +1203,51 @@
     assert content_block_start["id"] == "toolu_bdrk_013xRVejhv3ybmLEGCoZib2b"
 
 
+def test_streaming_chunk_with_text_and_empty_tool_calls_returns_text_delta():
+    """
+    Some OpenAI-compatible providers emit `tool_calls: []` on regular text chunks.
+
+    Empty tool_calls should be treated as no tool call so the Anthropic adapter
+    does not shadow text with an empty input_json_delta.
+    """
+    choices = [
+        StreamingChoices(
+            finish_reason=None,
+            index=0,
+            delta=Delta(
+                provider_specific_fields=None,
+                content="Hello from vLLM",
+                role="assistant",
+                function_call=None,
+                tool_calls=[],
+                audio=None,
+            ),
+            logprobs=None,
+        )
+    ]
+
+    adapter = LiteLLMAnthropicMessagesAdapter()
+
+    (
+        type_of_content,
+        content_block_delta,
+    ) = adapter._translate_streaming_openai_chunk_to_anthropic(choices=choices)
+
+    assert type_of_content == "text_delta"
+    assert content_block_delta["type"] == "text_delta"
+    assert content_block_delta["text"] == "Hello from vLLM"
+
+    (
+        block_type,
+        content_block_start,
+    ) = adapter._translate_streaming_openai_chunk_to_anthropic_content_block(
+        choices=choices
+    )
+
+    assert block_type == "text"
+    assert content_block_start == {"type": "text", "text": ""}
+
+
 # ============================================================================
 # Cache Control Transformation Tests
 # ============================================================================

diff --git a/tests/test_litellm/llms/azure/chat/test_azure_base_model_routing.py b/tests/test_litellm/llms/azure/chat/test_azure_base_model_routing.py
new file mode 100644
--- /dev/null
+++ b/tests/test_litellm/llms/azure/chat/test_azure_base_model_routing.py
@@ -1,0 +1,274 @@
+"""Tests for decoupling Azure deployment IDs from underlying model names.
+
+When users name their Azure deployment something non-standard (e.g. "my-deployment-id"),
+setting ``base_model`` should drive model-type detection (o-series, gpt-5,
+etc.) so the correct config, supported params, and param mapping are used.
+"""
+
+import pytest
+
+import litellm
+from litellm.llms.azure.chat.gpt_5_transformation import AzureOpenAIGPT5Config
+from litellm.llms.azure.chat.o_series_transformation import AzureOpenAIO1Config
+from litellm.utils import ProviderConfigManager, get_optional_params
+
+
+# ---------------------------------------------------------------------------
+# _get_azure_config — routes to the correct config based on base_model
+# ---------------------------------------------------------------------------
+class TestGetAzureConfigWithBaseModel:
+    """ProviderConfigManager._get_azure_config should use base_model for detection."""
+
+    def test_should_return_gpt5_config_when_base_model_is_gpt5(self):
+        config = ProviderConfigManager._get_azure_config(
+            model="my-deployment-id", base_model="azure/gpt-5.2"
+        )
+        assert isinstance(config, AzureOpenAIGPT5Config)
+
+    def test_should_return_o_series_config_when_base_model_is_o_series(self):
+        config = ProviderConfigManager._get_azure_config(
+            model="my-deployment-id", base_model="azure/o4-mini"
+        )
+        assert isinstance(config, AzureOpenAIO1Config)
+
+    def test_should_return_default_config_when_base_model_is_regular(self):
+        config = ProviderConfigManager._get_azure_config(
+            model="my-deployment-id", base_model="azure/gpt-4o"
+        )
+        assert type(config).__name__ == "AzureOpenAIConfig"
+
+    def test_should_fallback_to_model_when_base_model_is_none(self):
+        config = ProviderConfigManager._get_azure_config(
+            model="gpt-5.2", base_model=None
+        )
+        assert isinstance(config, AzureOpenAIGPT5Config)
+
+    def test_should_return_default_config_when_both_are_non_standard(self):
+        config = ProviderConfigManager._get_azure_config(
+            model="my-deployment-id", base_model=None
+        )
+        assert type(config).__name__ == "AzureOpenAIConfig"
+
+
+# ---------------------------------------------------------------------------
+# get_provider_chat_config — threads base_model through for Azure
+# ---------------------------------------------------------------------------
+class TestGetProviderChatConfigWithBaseModel:
+    """get_provider_chat_config should pass base_model to Azure config selection."""
+
+    def test_should_return_gpt5_config_for_custom_deployment_with_base_model(self):
+        from litellm.types.utils import LlmProviders
+
+        config = ProviderConfigManager.get_provider_chat_config(
+            model="my-deployment-id",
+            provider=LlmProviders.AZURE,
+            base_model="azure/gpt-5",
+        )
+        assert isinstance(config, AzureOpenAIGPT5Config)
+
+    def test_should_return_o_series_config_for_custom_deployment_with_base_model(self):
+        from litellm.types.utils import LlmProviders
+
+        config = ProviderConfigManager.get_provider_chat_config(
+            model="my-other-deployment",
+            provider=LlmProviders.AZURE,
+            base_model="azure/o3-mini",
+        )
+        assert isinstance(config, AzureOpenAIO1Config)
+
+
+# ---------------------------------------------------------------------------
+# get_supported_openai_params — base_model drives Azure param detection
+# ---------------------------------------------------------------------------
+class TestGetSupportedOpenAIParamsWithBaseModel:
+    """get_supported_openai_params should use base_model for Azure detection."""
+
+    def test_should_return_gpt5_params_for_custom_deployment_with_gpt5_base_model(
+        self,
+    ):
+        params = litellm.get_supported_openai_params(
+            model="my-deployment-id",
+            custom_llm_provider="azure",
+            base_model="azure/gpt-5",
+        )
+        assert params is not None
+        assert "reasoning_effort" in params
+        # gpt-5 maps max_tokens -> max_completion_tokens, verifying we got GPT-5 config
+        assert "max_completion_tokens" in params
+
+    def test_should_return_o_series_params_for_custom_deployment_with_o_series_base_model(
+        self,
+    ):
+        params = litellm.get_supported_openai_params(
+            model="my-other-deployment",
+            custom_llm_provider="azure",
+            base_model="azure/o4-mini",
+        )
+        assert params is not None
+        assert "reasoning_effort" in params
+
+    def test_should_return_regular_params_when_no_base_model(self):
+        """When base_model is not set and model is non-standard, default Azure config."""
+        params = litellm.get_supported_openai_params(
+            model="my-deployment-id",
+            custom_llm_provider="azure",
+        )
+        assert params is not None
+        # Default Azure config supports temperature
+        assert "temperature" in params
+
+
+# ---------------------------------------------------------------------------
+# get_optional_params — base_model drives Azure param mapping
+# ---------------------------------------------------------------------------
+class TestGetOptionalParamsWithBaseModel:
+    """get_optional_params should use base_model for Azure model-type detection."""
+
+    def test_should_map_max_tokens_for_custom_deployment_with_gpt5_base_model(self):
+        """A non-standard deployment name + gpt-5 base_model should map max_tokens -> max_completion_tokens."""
+        params = get_optional_params(
+            model="my-deployment-id",
+            custom_llm_provider="azure",
+            max_tokens=100,
+            base_model="azure/gpt-5",
+        )
+        assert params.get("max_completion_tokens") == 100
+        assert "max_tokens" not in params
+
+    def test_should_keep_max_tokens_for_custom_deployment_without_base_model(self):
+        """A non-standard deployment name without base_model should use default Azure config."""
+        params = get_optional_params(
+            model="my-deployment-id",
+            custom_llm_provider="azure",
+            max_tokens=100,
+            api_version="2024-05-01-preview",
+        )
+        # Default AzureOpenAIConfig keeps max_tokens as-is (or maps based on api_version)
+        assert "max_tokens" in params or "max_completion_tokens" in params
+
+    def test_should_support_reasoning_effort_for_custom_deployment_with_o_series_base_model(
+        self,
+    ):
+        """A non-standard deployment name + o-series base_model should accept reasoning_effort."""
+        params = get_optional_params(
+            model="my-other-deployment",
+            custom_llm_provider="azure",
+            reasoning_effort="low",
+            base_model="azure/o4-mini",
+        )
+        assert params.get("reasoning_effort") == "low"
+
+    def test_should_reject_temperature_for_custom_deployment_with_gpt5_base_model(
+        self,
+    ):
+        """A non-standard deployment + gpt-5 base_model should reject temperature."""
+        with pytest.raises(litellm.UnsupportedParamsError):
+            get_optional_params(
+                model="my-deployment-id",
+                custom_llm_provider="azure",
+                temperature=0.5,
+                base_model="azure/gpt-5",
+            )
+
+
+# ---------------------------------------------------------------------------
+# Backward compatibility — existing patterns still work
+# ---------------------------------------------------------------------------
+class TestBackwardCompatibility:
+    """Existing model-name-based and prefix-based patterns must keep working."""
+
+    def test_should_detect_gpt5_from_model_name(self):
+        config = ProviderConfigManager._get_azure_config(model="gpt-5.2")
+        assert isinstance(config, AzureOpenAIGPT5Config)
+
+    def test_should_detect_gpt5_from_gpt5_series_prefix(self):
+        config = ProviderConfigManager._get_azure_config(
+            model="gpt5_series/my-deployment"
+        )
+        assert isinstance(config, AzureOpenAIGPT5Config)
+
+    def test_should_detect_o_series_from_model_name(self):
+        config = ProviderConfigManager._get_azure_config(model="o4-mini")
+        assert isinstance(config, AzureOpenAIO1Config)
+
+    def test_should_detect_o_series_from_o_series_prefix(self):
+        config = ProviderConfigManager._get_azure_config(model="o_series/my-deployment")
+        assert isinstance(config, AzureOpenAIO1Config)
+
+    def test_should_handle_gpt5_chat_model_correctly(self):
+        """gpt-5-chat models should NOT be routed to GPT-5 config."""
+        config = ProviderConfigManager._get_azure_config(model="gpt-5-chat")
+        assert type(config).__name__ == "AzureOpenAIConfig"
+
+    def test_base_model_overrides_model_detection(self):
+        """base_model should take priority over model for type detection."""
+        # model looks like o-series, but base_model says gpt-5
+        config = ProviderConfigManager._get_azure_config(
+            model="o3-mini", base_model="azure/gpt-5.2"
+        )
+        assert isinstance(config, AzureOpenAIGPT5Config)
+
+
+# ---------------------------------------------------------------------------
+# Deep config method awareness — base_model flows into config internals
+# ---------------------------------------------------------------------------
+class TestBaseModelFlowsIntoConfigInternals:
+    """base_model should be used by config internal methods (e.g. is_model_gpt_5_2_model)."""
+
+    def test_should_support_logprobs_for_prefixed_deployment_with_gpt52_base_model(
+        self,
+    ):
+        """Deployment 'my-gpt-5.2' with base_model='azure/gpt-5.2' should support logprobs."""
+        params = litellm.get_supported_openai_params(
+            model="gpt5_series/my-gpt-5.2",
+            custom_llm_provider="azure",
+            base_model="azure/gpt-5.2",
+        )
+        assert params is not None
+        assert "logprobs" in params
+        assert "top_logprobs" in params
+
+    def test_should_support_logprobs_for_plain_deployment_with_gpt52_base_model(self):
+        """Deployment 'my-deployment-id' with base_model='azure/gpt-5.2' should support logprobs."""
+        params = litellm.get_supported_openai_params(
+            model="my-deployment-id",
+            custom_llm_provider="azure",
+            base_model="azure/gpt-5.2",
+        )
+        assert params is not None
+        assert "logprobs" in params
... diff truncated: showing 800 of 1138 lines